Matthew Wilcox | c94c2ac | 2015-09-08 14:58:40 -0700 | [diff] [blame] | 1 | #ifndef _LINUX_DAX_H |
| 2 | #define _LINUX_DAX_H |
| 3 | |
| 4 | #include <linux/fs.h> |
| 5 | #include <linux/mm.h> |
Jan Kara | 4f62293 | 2016-05-12 18:29:17 +0200 | [diff] [blame] | 6 | #include <linux/radix-tree.h> |
Matthew Wilcox | c94c2ac | 2015-09-08 14:58:40 -0700 | [diff] [blame] | 7 | #include <asm/pgtable.h> |
| 8 | |
Christoph Hellwig | a254e56 | 2016-09-19 11:24:49 +1000 | [diff] [blame] | 9 | struct iomap_ops; |
Dan Williams | 6568b08 | 2017-01-24 18:44:18 -0800 | [diff] [blame] | 10 | struct dax_device; |
| 11 | struct dax_operations { |
| 12 | /* |
| 13 | * direct_access: translate a device-relative |
| 14 | * logical-page-offset into an absolute physical pfn. Return the |
| 15 | * number of pages available for DAX at that pfn. |
| 16 | */ |
| 17 | long (*direct_access)(struct dax_device *, pgoff_t, long, |
| 18 | void **, pfn_t *); |
Dan Williams | 5d61e43 | 2017-06-27 13:06:22 -0700 | [diff] [blame] | 19 | /* copy_from_iter: required operation for fs-dax direct-i/o */ |
Dan Williams | 0aed55a | 2017-05-29 12:22:50 -0700 | [diff] [blame] | 20 | size_t (*copy_from_iter)(struct dax_device *, pgoff_t, void *, size_t, |
| 21 | struct iov_iter *); |
Dan Williams | 3c1cebf | 2017-05-29 12:58:19 -0700 | [diff] [blame] | 22 | /* flush: optional driver-specific cache management after writes */ |
| 23 | void (*flush)(struct dax_device *, pgoff_t, void *, size_t); |
Dan Williams | 6568b08 | 2017-01-24 18:44:18 -0800 | [diff] [blame] | 24 | }; |
Christoph Hellwig | a254e56 | 2016-09-19 11:24:49 +1000 | [diff] [blame] | 25 | |
Dan Williams | 6e0c90d | 2017-06-26 21:28:41 -0700 | [diff] [blame] | 26 | extern struct attribute_group dax_attribute_group; |
| 27 | |
Dan Williams | ef510424 | 2017-05-08 10:55:27 -0700 | [diff] [blame] | 28 | #if IS_ENABLED(CONFIG_DAX) |
| 29 | struct dax_device *dax_get_by_host(const char *host); |
| 30 | void put_dax(struct dax_device *dax_dev); |
| 31 | #else |
| 32 | static inline struct dax_device *dax_get_by_host(const char *host) |
| 33 | { |
| 34 | return NULL; |
| 35 | } |
| 36 | |
| 37 | static inline void put_dax(struct dax_device *dax_dev) |
| 38 | { |
| 39 | } |
| 40 | #endif |
| 41 | |
Dan Williams | f5705aa8c | 2017-05-13 16:31:05 -0700 | [diff] [blame] | 42 | int bdev_dax_pgoff(struct block_device *, sector_t, size_t, pgoff_t *pgoff); |
| 43 | #if IS_ENABLED(CONFIG_FS_DAX) |
| 44 | int __bdev_dax_supported(struct super_block *sb, int blocksize); |
| 45 | static inline int bdev_dax_supported(struct super_block *sb, int blocksize) |
| 46 | { |
| 47 | return __bdev_dax_supported(sb, blocksize); |
| 48 | } |
| 49 | |
| 50 | static inline struct dax_device *fs_dax_get_by_host(const char *host) |
| 51 | { |
| 52 | return dax_get_by_host(host); |
| 53 | } |
| 54 | |
| 55 | static inline void fs_put_dax(struct dax_device *dax_dev) |
| 56 | { |
| 57 | put_dax(dax_dev); |
| 58 | } |
| 59 | |
| 60 | #else |
| 61 | static inline int bdev_dax_supported(struct super_block *sb, int blocksize) |
| 62 | { |
| 63 | return -EOPNOTSUPP; |
| 64 | } |
| 65 | |
| 66 | static inline struct dax_device *fs_dax_get_by_host(const char *host) |
| 67 | { |
| 68 | return NULL; |
| 69 | } |
| 70 | |
| 71 | static inline void fs_put_dax(struct dax_device *dax_dev) |
| 72 | { |
| 73 | } |
| 74 | #endif |
| 75 | |
Dan Williams | 7b6be84 | 2017-04-11 09:49:49 -0700 | [diff] [blame] | 76 | int dax_read_lock(void); |
| 77 | void dax_read_unlock(int id); |
Dan Williams | c1d6e82 | 2017-01-24 23:02:09 -0800 | [diff] [blame] | 78 | struct dax_device *alloc_dax(void *private, const char *host, |
| 79 | const struct dax_operations *ops); |
Dan Williams | c1d6e82 | 2017-01-24 23:02:09 -0800 | [diff] [blame] | 80 | bool dax_alive(struct dax_device *dax_dev); |
| 81 | void kill_dax(struct dax_device *dax_dev); |
| 82 | void *dax_get_private(struct dax_device *dax_dev); |
Dan Williams | b068626 | 2017-01-26 20:37:35 -0800 | [diff] [blame] | 83 | long dax_direct_access(struct dax_device *dax_dev, pgoff_t pgoff, long nr_pages, |
| 84 | void **kaddr, pfn_t *pfn); |
Dan Williams | 7e026c8 | 2017-05-29 12:57:56 -0700 | [diff] [blame] | 85 | size_t dax_copy_from_iter(struct dax_device *dax_dev, pgoff_t pgoff, void *addr, |
| 86 | size_t bytes, struct iov_iter *i); |
Dan Williams | abebfbe | 2017-05-29 13:02:52 -0700 | [diff] [blame] | 87 | void dax_flush(struct dax_device *dax_dev, pgoff_t pgoff, void *addr, |
| 88 | size_t size); |
Dan Williams | 6e0c90d | 2017-06-26 21:28:41 -0700 | [diff] [blame] | 89 | void dax_write_cache(struct dax_device *dax_dev, bool wc); |
Vivek Goyal | 273752c | 2017-07-26 09:35:09 -0400 | [diff] [blame] | 90 | bool dax_write_cache_enabled(struct dax_device *dax_dev); |
Dan Williams | 7b6be84 | 2017-04-11 09:49:49 -0700 | [diff] [blame] | 91 | |
Ross Zwisler | fa28f72 | 2016-11-08 11:33:35 +1100 | [diff] [blame] | 92 | /* |
Ross Zwisler | 642261a | 2016-11-08 11:34:45 +1100 | [diff] [blame] | 93 | * We use lowest available bit in exceptional entry for locking, one bit for |
| 94 | * the entry size (PMD) and two more to tell us if the entry is a huge zero |
| 95 | * page (HZP) or an empty entry that is just used for locking. In total four |
| 96 | * special bits. |
| 97 | * |
| 98 | * If the PMD bit isn't set the entry has size PAGE_SIZE, and if the HZP and |
| 99 | * EMPTY bits aren't set the entry is a normal DAX entry with a filesystem |
| 100 | * block allocation. |
Ross Zwisler | fa28f72 | 2016-11-08 11:33:35 +1100 | [diff] [blame] | 101 | */ |
Ross Zwisler | 642261a | 2016-11-08 11:34:45 +1100 | [diff] [blame] | 102 | #define RADIX_DAX_SHIFT (RADIX_TREE_EXCEPTIONAL_SHIFT + 4) |
Jan Kara | e804315 | 2016-05-12 18:29:16 +0200 | [diff] [blame] | 103 | #define RADIX_DAX_ENTRY_LOCK (1 << RADIX_TREE_EXCEPTIONAL_SHIFT) |
Ross Zwisler | 642261a | 2016-11-08 11:34:45 +1100 | [diff] [blame] | 104 | #define RADIX_DAX_PMD (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 1)) |
| 105 | #define RADIX_DAX_HZP (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 2)) |
| 106 | #define RADIX_DAX_EMPTY (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 3)) |
Ross Zwisler | fa28f72 | 2016-11-08 11:33:35 +1100 | [diff] [blame] | 107 | |
Ross Zwisler | 642261a | 2016-11-08 11:34:45 +1100 | [diff] [blame] | 108 | static inline unsigned long dax_radix_sector(void *entry) |
| 109 | { |
| 110 | return (unsigned long)entry >> RADIX_DAX_SHIFT; |
| 111 | } |
| 112 | |
| 113 | static inline void *dax_radix_locked_entry(sector_t sector, unsigned long flags) |
| 114 | { |
| 115 | return (void *)(RADIX_TREE_EXCEPTIONAL_ENTRY | flags | |
| 116 | ((unsigned long)sector << RADIX_DAX_SHIFT) | |
| 117 | RADIX_DAX_ENTRY_LOCK); |
| 118 | } |
Jan Kara | e804315 | 2016-05-12 18:29:16 +0200 | [diff] [blame] | 119 | |
Ross Zwisler | 11c59c9 | 2016-11-08 11:32:46 +1100 | [diff] [blame] | 120 | ssize_t dax_iomap_rw(struct kiocb *iocb, struct iov_iter *iter, |
Christoph Hellwig | 8ff6daa | 2017-01-27 23:20:26 -0800 | [diff] [blame] | 121 | const struct iomap_ops *ops); |
Dave Jiang | c791ace | 2017-02-24 14:57:08 -0800 | [diff] [blame] | 122 | int dax_iomap_fault(struct vm_fault *vmf, enum page_entry_size pe_size, |
| 123 | const struct iomap_ops *ops); |
Jan Kara | ac401cc | 2016-05-12 18:29:18 +0200 | [diff] [blame] | 124 | int dax_delete_mapping_entry(struct address_space *mapping, pgoff_t index); |
Jan Kara | c6dcf52 | 2016-08-10 17:22:44 +0200 | [diff] [blame] | 125 | int dax_invalidate_mapping_entry_sync(struct address_space *mapping, |
| 126 | pgoff_t index); |
Jan Kara | ac401cc | 2016-05-12 18:29:18 +0200 | [diff] [blame] | 127 | void dax_wake_mapping_entry_waiter(struct address_space *mapping, |
Ross Zwisler | 63e95b5 | 2016-11-08 11:32:20 +1100 | [diff] [blame] | 128 | pgoff_t index, void *entry, bool wake_all); |
Dan Williams | d1a5f2b4 | 2016-01-28 20:25:31 -0800 | [diff] [blame] | 129 | |
| 130 | #ifdef CONFIG_FS_DAX |
Dan Williams | cccbce6 | 2017-01-27 13:31:42 -0800 | [diff] [blame] | 131 | int __dax_zero_page_range(struct block_device *bdev, |
| 132 | struct dax_device *dax_dev, sector_t sector, |
Christoph Hellwig | 679c8bd | 2016-05-09 10:47:04 +0200 | [diff] [blame] | 133 | unsigned int offset, unsigned int length); |
Dan Williams | d1a5f2b4 | 2016-01-28 20:25:31 -0800 | [diff] [blame] | 134 | #else |
Christoph Hellwig | 679c8bd | 2016-05-09 10:47:04 +0200 | [diff] [blame] | 135 | static inline int __dax_zero_page_range(struct block_device *bdev, |
Dan Williams | cccbce6 | 2017-01-27 13:31:42 -0800 | [diff] [blame] | 136 | struct dax_device *dax_dev, sector_t sector, |
| 137 | unsigned int offset, unsigned int length) |
Christoph Hellwig | 679c8bd | 2016-05-09 10:47:04 +0200 | [diff] [blame] | 138 | { |
| 139 | return -ENXIO; |
| 140 | } |
Dan Williams | d1a5f2b4 | 2016-01-28 20:25:31 -0800 | [diff] [blame] | 141 | #endif |
| 142 | |
Ross Zwisler | 642261a | 2016-11-08 11:34:45 +1100 | [diff] [blame] | 143 | #ifdef CONFIG_FS_DAX_PMD |
| 144 | static inline unsigned int dax_radix_order(void *entry) |
| 145 | { |
| 146 | if ((unsigned long)entry & RADIX_DAX_PMD) |
| 147 | return PMD_SHIFT - PAGE_SHIFT; |
| 148 | return 0; |
| 149 | } |
Ross Zwisler | 642261a | 2016-11-08 11:34:45 +1100 | [diff] [blame] | 150 | #else |
| 151 | static inline unsigned int dax_radix_order(void *entry) |
| 152 | { |
| 153 | return 0; |
| 154 | } |
Ross Zwisler | 642261a | 2016-11-08 11:34:45 +1100 | [diff] [blame] | 155 | #endif |
Dave Jiang | 11bac80 | 2017-02-24 14:56:41 -0800 | [diff] [blame] | 156 | int dax_pfn_mkwrite(struct vm_fault *vmf); |
Matthew Wilcox | c94c2ac | 2015-09-08 14:58:40 -0700 | [diff] [blame] | 157 | |
Ross Zwisler | f9fe48b | 2016-01-22 15:10:40 -0800 | [diff] [blame] | 158 | static inline bool dax_mapping(struct address_space *mapping) |
| 159 | { |
| 160 | return mapping->host && IS_DAX(mapping->host); |
| 161 | } |
Ross Zwisler | 7f6d5b5 | 2016-02-26 15:19:55 -0800 | [diff] [blame] | 162 | |
| 163 | struct writeback_control; |
| 164 | int dax_writeback_mapping_range(struct address_space *mapping, |
| 165 | struct block_device *bdev, struct writeback_control *wbc); |
Matthew Wilcox | c94c2ac | 2015-09-08 14:58:40 -0700 | [diff] [blame] | 166 | #endif |